Modeling Public and Private Real Estate Risk · with such high reliance on alternative asset...

Modeling Public and Private Real Estate Risk

Richard Gold Northfield Webinar March 2014

http://www.google.com/url?sa=i&rct=j&q=&esrc=s&frm=1&source=images&cd=&cad=rja&docid=BLjRfL-U4eWKtM&tbnid=iueqjhHN1iTmgM:&ved=0CAUQjRw&url=http://japho.com/how-to-build-a-brick-wall/&ei=2uyUUfTXKIeZ0QH23oHgCQ&bvm=bv.46471029,d.dmg&psig=AFQjCNHnbG5Du3YfLYdYGoCE_xTbO4enKw&ust=1368800822494152

www.northinfo.com Slide 2

Northfield’s Risk Management Principals

• Risk is about the future not the past. Every aspect of the asset owner’s portfolio must be examined in an explic itly forw ard-look ing modeling framew ork over multiple time horizons. Simply relying on historic measures of fund volatility or tracking error is just not good enough.

• 100% of all assets in a fund must be reviewed regularly in granular detail. In today’s world with such high reliance on alternative asset classes, dealing only with risk at the asset allocation level is grossly inadequate.

• If you ow n an asset, you need to understand how it impac ts risk at both the asset c lass and portfolio levels. If you don’t, then you might have bought a pig in a poke.


Introduction

• In the spirit of the previous slide, I will cover how Northfield models risk for: – U.S. and global REITs – The critical distinction between public and private real estate investment

vehicles and why they need to be treated differently – Private equity real estate – Briefly look at the question of leverage and its impact on property risk – I will not be covering CMBS and single family mortgage pools but leave those

for a later discussion

Risk = Uncertainty of Return = Standard Deviation


Main Concepts for Today

• Investment practitioners rely on a decomposition of portfolio risk into factors to guide investment decisions

• While investment risk is usually unambiguous, it is not necessarily the case for real estate:

– While total risk for publically-traded REITs may be easily determined…. – Risk for appraisal-based private equity real estate is not

• Biased risk estimators lead to a host of risk decomposition and allocation problems – This is an issue for private equity real estate and appraisals


Implicit Decomposition in Factor Specification

• All factor models rely on a simple linear representation of asset (or portfolio) returns

Rt = Σi=1 to n BiFit + εt

Rt = the asset return in period t

Bi = the factor exposure to factor i Fit = the return to factor i in period t εt = the error term in period t

In times series models we observe the F values and statistically estimate the B values. In “fundamental” models we observe the B values and statistically estimate the F values. In blind factor (PCA) models we jointly estimate both.


Implicit Decomposition in Factor Specifications

• Let’s assume we have two categories of factors called “red factors” and “blue factors”. We could write such a model as to have G red factors and H blue factors

Rt = Σi=1 to g BiFit + Σi=g+1 to g+h BiFit +εt

If we had a reason to do so, we could define the blue factors as being “net” of the

influence of the red factors. For example, if inflation and interest rates were both factors in the model, we might choose to put inflation in the red group and redefine “nominal interest rates” as “interest rates net of the effect of inflation and other red factors (e.g. real interest rates).


Implicit Decomposition in Factor Specifications

• We accomplish this structuring of the decomposition by using a two step estimation procedure

Rt = Σi=1 to g BiFit + ζt ζt = Σi=g+1 to g+h BiFit +εt

ζt = the residual return at time t net of red factors only

Since we have defined and estimated the blue factors net of the red factors the risk decomposition will naturally allocate more risk to the red factors and less to the blue factors


Why do “staged” Model Estimation?

• You have more data for some factors than others. – You have a universe of 1000 stocks broken into 50 industries – You will have 1000 data points for estimating the return to a factor like P/E or size, but

only an average of 20 data points to estimate the return to a particular industry group.

• You have two or more factors that are highly correlated – Statistical estimation procedures often produce unstable results when independent

variables are correlated. – By defining one factor net of another correlated factor, we structurally remove their

natural correlation

• You have particular strategies where it makes sense – There has been a long debate about whether countries/regions or sectors/industries are

more important to global equity portfolios. – In fact, that is what we do in the Northfield REIT models


Northfield REIT Risk Models

• In fact this two step approach is exactly how Northfield estimates both its Global and U.S. REIT models

• In both the U.S. and Global REIT Models the regional effect is thought to dominate the property type effect once the global industry effect is taken into account

– A rising tide lifts all boats – If Asia is doing well – all properties in the region tend to do well

– For many years the dearth of REITs in many locations made it difficult to target land uses

– Common legal structure of REITs adds to regional homogeneity and behavior • In the US investors allocate by land use and then by geography

– Larger REITs traditionally rewarded: – Economies of scale: Lower overhead and financing costs – Portfolio diversification


160+ REITS Weekly data

• Model estimated over last 104 weeks

9 Industry Indices • Diversified • Mortgage • Healthcare: • Industrial • Office • Hotels • Residential • Retail • Self-Storage

9 Geographic Regions • East North Central • Southwest • West North Central • Mountain • Northwest • Pacific • Mideast • Southeast • Unclassified

Equally-Weight U.S. Real Estate Index Size Index

Northfield U.S. REIT Model

Model Specification


• The U.S. Model is estimated using a two step process 1. STEP I: Regress individual REIT returns against:

a) An equal-weighted index of all REITs in Northfield’s universe b) A “pure” or residual return industry sector series. This series is the return of each

industry sector factor that cannot be explained by the market index factor. It is obtained by running a simple regression of each industry sector index against the U.S. REIT index and obtaining its residual.

c) A size factor equal to the difference in equally weighted average returns between the largest and the smallest REIT companies in the top and bottom quartiles

2. STEP II: Residuals from Step 1 are regressed against nine regional factors based on the REIT’s property exposure in a region

• The standard deviation of the residuals from Step 2 is security-specific risk • A REIT can only participate in one industry sector • Mortgage REITs tend to be unstable so “caveat emptor”

Northfield U.S. REIT Model (con’t)


440+ REITS Weekly data

• Model estimated over last 104 weeks

10 Industry Indices • Diversified • Healthcare • Self-Storage • Industrial • Office • Residential • Retail • Lodging/Resorts • Specialty • Industrial/Office Mixed

3 Geographic Regions • North America • Europe • Asia

Global Real Estate Index Size Index 5 Statistical Factors (PCA) Currency Beta (Return Relative to the U.S.

Dollar)

Northfield Global REIT Model

Model Specification


• The Global Model uses a three step estimation process as well as a final currency beta representing the return of each REIT’s currency against U.S. Dollar .

1. STEP I: Individual REIT returns are regressed against the FTSE EPRA/NAREIT Index 2. STEP II: The residuals of the first equation are regressed against:

a) A “pure” or residual return sector series. This series is the return of each sector factor that cannot be explained by the market index factor. It is obtained by running a simple regression of each FTSE EPRA/NAREIT sector index against FTSE EPRA/NAREIT global index and obtaining its residual.

b) A size variable obtained by creating two indices representing top and bottom quartiles of the global REIT market. Size factor returns are equal to top quartile minus bottom quartile.

c) Regional dummy variables 3. The residuals from the second regression are then used to create the covariance matrix that

is then employed in the final step: the estimation of 5 Statistical Factors using Principal Components Analysis (PCA)

a) Each security’s beta sensitivity to each of the five statistical factors is then estimated 4. Each currency factor represents the return of each currency against U.S. Dollar.

Northfield Global REIT Model (con’t)

10


• The security-specific risk for each security is calculated by taking the standard deviation of the residual return remaining after the estimation of the statistical factors

• No causality is assigned to the statistical factors. • They are transitory • Users can assign their own attribution but Northfield does not

• REITs can only be in one region • Size variable is a market momentum variable

Northfield Global REIT Model (con’t)


Factor PortExp BenchExp ActiveExp FactorVar VarContr REIT Market 0.882 0.944 -0.062 338.632 1.665 East North Central 0.044 0.074 -0.030 794.775 0.832 Residential 0.088 0.056 0.031 214.993 0.569 Northeast 0.253 0.183 0.070 91.935 0.612

Factor Tracking Variance 5.870 Stock Specific Tracking Variance 11.397 Total Tracking Variance 17.267 Tracking Error 4.155 Total Risk of Portfolio 18.594 Total Risk of Benchmark 19.708 R-Squared 0.957

U.S. Model – Risk Decomposition Report Random Portfolio of 20 Equally-Weighted REITs

Data Through January 31, 2014


Private Equity Real Estate • “Fama-French” is not common jargon in real estate investment departments

• Unlisted investment experts tend to think of risk in terms of first distributional moment, i.e. return, not

volatility

• “Location, Location…” is not an obvious quantity nor is it constant over time

• Current indices (NCREIF, IPD, etc.) do not use observable prices: – Appraisal-driven and therefore suffer from appraisal bias

– Smoothed, dampened: exhibit serial correlation – Repeat Sales Indices: small sample sizes & little property-level info

• Current indices: – Good for:

– Historic trends and long-term absolute returns (See: Cheng, Lin, & Liu 2011) – Not good for:

– Short-term & mid-term analysis, asset allocation – Uncertainty of future returns (risk) – Risk components and its contribution at the property and portfolio levels


Public Versus Private Property Returns

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

1994:1 2000:3 2007:1 2013:3

QUA

RTER

LY R

ETU

RN

NCREIF VERSUS NAREIT 1994Q1 - 2013Q3

NAREIT

NCREIF

Source: NCREIF & NAREIT


Appraisal Bias

0 0.5 1

LAG 1

LAG 2

LAG 3

LAG 4

LAG 5

NCREIF

NCREIF Open-End Funds

RETURN PERSISTENCE REMAINS AN ISSUE LAGGED QUARTERLY BETAS

2003Q4 – 2013Q3

Source: NCREIF


Correcting for Autocorrelation

Standard Deviation - Last 20 Quarters of Available Data

3.9%

5.8%

5.7%

7.6%

8.7%

14.1%

21.3%

20.7%

26.5%

30.2%

NCREIF

NCREIF OPEN-END FUNDS

TOWNSEND CORE

TOWNSEND VALUE ADDED

TOWNSEND OPPORTUNISTIC

ADJUSTED

UNADJUSTED

Source: NCREIF

No Leverage

www.northinfo.com Slide 20 20

Northfield’s Private Property Risk Approach

• A “bottom-up” property-by-property model that is not appraisal-based • Each property is a “composite asset” with three attributes:

– Risks based on “steady-state” cash flow assumptions for existing and expected leases • Uses lease structure, renewal, credit quality of tenants, vacancy dynamics, revenue

and expense schedules – Risks related to mortgage financing (if any)

• Takes into consideration floating rate, fixed rate, interest-only, balloon clauses, prepayment behavior, etc.

– Risks of future fluctuations in market rents • Takes into consideration the combined impact of lease rollover, vacancy, renewal, and

market volatility of rents • Each component has risk exposures to common risk factors plus

idiosyncratic risks • A second less “granular” approach employing “generic” properties can be

employed • A third but even less informative approach is to use a basket of “like kind”

REITS weighted by geography and property type


Private Real Estate Model Structure MORTGAGE FINANCING

(SHORT)

STEADY STATE CASH FLOW

(LONG)

RENT VOLATILITY

TIME VALUE OF MONEY

CREDIT RISK

CHANGE IN RENT

EE MODEL

PROPERTY/ PORTFOLIO RISK

RISK FACTORS RISK FACTORS RISK FACTORS


Components of Property-Level Risk

Rent Risk Interest Rate Risk

Credit Risk

Asset Specific

Risk

TOTAL RISK


Model Results – A Sample Portfolio: 15% Gearing

23

PORTFOLIO PROFILE Metro Apartment Office Industrial Retail

Berlin 1

Budapest 1

Frankfurt 1

Rome 2

London 1 2 1

Marseilles 1

Amsterdam 1

Paris 1

Bucharest 1

Stockholm 1


Model Results – Everything Everywhere Model

24

Portfolio Risk Report / Euro-Based Investor

Factor PortExp BenchExp ActiveExp FactorVar VarContr S B WORLD GOVT BOND INDEX -0.02 0 -0.02 44.41 -0.24 OIL PRICES IN USD 0.00 0 0.00 746.77 -0.11 TECHNOLOGY&HEALTH SECTOR 0.01 0 0.01 161.43 0.23 VALUE/GROWTH -0.05 0 -0.05 6.47 0.28 INTEREST RATE SENSITIVE SECTR 0.01 0 0.01 204.85 0.52 CONSUMER SECTOR 0.05 0 0.05 134.21 2.16 ENGLISH-SPEAKING COUNTRIES 0.05 0 0.05 213.65 2.41 SWEDISH KRONA 0.14 0 0.14 47.22 3.77 CONTINENTAL EUROPE 0.07 0 0.07 208.25 5.57 UK POUND 0.46 0 0.46 85.70 26.14 TREASURY CURVE FACTOR1 -26.68 0 -26.68 0.30 251.95 TREASURY CURVE FACTOR2 -235.01 0 -235.01 0.00 199.02 TREASURY CURVE FACTOR3 -1735.99 0 -1735.99 0.00 -120.53 Factor Tracking Variance 371.33 Stock Specific Tracking Variance 3.99 Total Tracking Variance 375.32 Tracking Error 19.37 Total Risk of Portfolio 19.37


Model Results – Geospatial Risk

25

Country Risk (All Leverage Removed)

Country Risk Property Count Land Use

Romania 15.9 1 Residential

Sweeden 16.4 1 Retail

Germany 17.6 2 Residential, Office

France 19.8 2 Retail, Office

Netherlands 19.9 1 Industrial

U.K. 20.3 4 Office(2), Industrial, Residential

Italy 20.4 2 Industrial(2)

Hungary 20.7 1 Retail


Expected Annualized Risk by Asset Class

26

- 12 Month Ex Ante as of November 2013 per Northfield’s EE model - Bond Index: Barclay’s U.S. Aggregate Bond Index -VGK: Vanguard European Equity Index (Approximately 506 Holdings) -REIT Portfolio: 93 European Equity REITs

0%

5%

10%

15%

20%

25%

Granular RE Port Generic RE Port "REIT" RE Port VGK Barclay's Bond Index

European REITs

Euro-Based Investor – November 2013

Last 5 Years

Ex Ante Risk


Effects of Leverage on Equity Risk – Single Property

27

0

20

40

60

80

100

120

140

160

0.2 0.4 0.6 0.95 1.5

Stan

dard

Dev

iatio

n

NYC Office Building: Risk & LTV

Debt as a Percent of Equity

For every percent increase in leverage over 60%, risk increases by over 3.2% until 95% LTV.

Leverage > 95% results in little chance of covering debt resulting in lower risk: no equity to loose and insufficient cash flow just waiting to hand back keys.


Assuming You Are Still Awake……

Slide 28


Conclusions

• REIT risk models within a factor model framework are straightforward and exhibit robust results

• Northfield’s multistage REIT models show that individual REITs have exposures to: – REIT Benchmark – Property type – Geographic region – Market momentum conditions – REITs in our Global Model have additional exposures to:

– Transitory Statistical Effects – Currency exposures if held in conjunction with other global REITs and/or held

outside of an investor’s local currency • Output from the models can be optimized and investors can actively manage their positions

once they understand their exposures.


Conclusions (con’t)

• Private equity real estate is a much more complex animal • Appraisal-based risk metrics are biased and cannot be used “out of the box”

– Even after correcting for “serial correlation” appraisal-based series have a host of issues at the asset level, asset allocation, and for ERM

• However, it is possible to transform a property into a composite asset: – Risks based on “steady-state” cash flow – Risks related to mortgage financing (if any) – Risks of future fluctuations in market rents – Each component has risk exposures to common risk factors plus idiosyncratic risks

• While interest rate risk is clearly an importantly factor, private equity real estate risk is also influenced by a host of other factors including credit quality, lease duration, and industry and market factors.

• if you ow n or are considering ow ning an asset, you should know as much about the investment as possible. Otherw ise you are tak ing on a different risk profile than you initially assumed.

Modeling Public and Private Real Estate Risk · with such high reliance on alternative asset...

Documents

Transcript of Modeling Public and Private Real Estate Risk · with such high reliance on alternative asset...